MAINTENANCE SYSTEM FOR AI KNOWLEDGE BASES

THIS PROPOSAL IS FOR THE DESIGN OF A SHELL CALLED SKRAM WHICH, WHEN IMPLEMENTED, WILL BE A SOFTWARE TOOL WHICH CAN CONSIDERABLY REDUCE THE DEVELOPMENT AND MAINTENANCE PROBLEMS ASSOCIATED WITH REAL-TIME RTIFICIAL INTELLIGENCE SYSTEMS. THE PROPOSED SHELL WILL INCORPORATE THE RESULTS OF OVER A DECADE OF RESEARCH INTO METHODS FOR DEVELOPING, TESTING, AND MAINTAINING KNOWLEDGE BASED SYSTEMS WHICH FUNCTION IN REAL-TIME ENVIRONMENTS. SKRAM WILL BE BASED ON THE USE OF EVIDENCE FLOW GRAPHS (EFGS) WHICH ARE A COMMON DATA FLOW REPRESENTATION FOR INTELLIGENT SYSTEMS. USERS WILL BE ABLE TO DEVELOP AND MAINTAIN THEIR SYSTEMS DIRECTLY AS EFGS OR TO IMPORT RULES OR OTHER KNOWLEDGE BASES AND HAVE THESE AUTOMATICALLY TRANSLATED TO EFGS. SKRAM WILL HAVE THE CAPABILITY TO TRANSLATE EFGS INTO ADA CODE MODULES AND THEN TO LINK THESE WITH THE ACTIVATION FRAMEWORK EXECUTION ENVIRONMENT TO PRODUCE A RUN-TIME SYSTEM WHICH IS CAPABLE OF EXECUTION IN REAL-TIME ON A PARALLEL PROCESOR SYSTEM. SKRAM WILL HAVE EXTENSIVE DEBUGGING AND TESTING FEATURES. IT WILL ALSO INCORPORATE THE USE OF A CONSTRAINT BASED TEST LANGUAGE FOR AUTOMATIC TEST GENERATION AND EVALUATION TO SUPPORT PERFORMANCE VALIDATION AND MAINTENANCE VERIFICATION.